Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.upflow.io/llms.txt

Use this file to discover all available pages before exploring further.

Estimated setup time: ~8 minutes. Use this guide to set up the Upflow MCP Server in your AI client (the example below uses Claude Code) so it can securely query Upflow on your behalf.

​
Limitations

The current version does not support remote AI clients such as Claude Desktop or ChatGPT, due to authentication constraints.
  • MCP authentication relies on Upflow API credentials, which grant read-only admin access across the entire organization.
  • API credentials are scoped to a single organization. To connect an AI client to multiple organizations, you’ll need to repeat the setup for each one.

​
Prerequisites

  • Connect with your CSM so we can confirm:
    • The current status of the MCP Server and that we can support your use case
    • That your technical team understands the authentication constraints
  • Once confirmed, the Upflow team will activate the feature so you can start using it.

​
Step 1 – Collect your connection parameters

  1. In the Upflow app, go to: Organization settings → Apps → API.
  2. Copy the following values:
    • Host
    • Key
    • Secret
  3. For the MCP host, take the Host value and append mcp at the end.
    • Example: https://api.upflow.io → https://api.upflow.io/mcp
  4. For User email, use:
    • Your own email, or
    • Any email belonging to a member of the same organization.
You should now have:
  • Host (with /mcp at the end)
  • Key
  • Secret
  • User_email

​
Step 2 – Add Upflow MCP to your AI client

Below are examples for different clients. In each case, replace the placeholders with the values from Step 1.

​
Example with Claude Code (CLI)

If you don’t have Claude Code yet, use https://code.claude.com/docs/en/quickstart to get started. You will need a Claude Code Sub. Run this command in your terminal: 
curl -fsSL <https://claude.ai/install.sh> | bash
This will prompt you to add Claude Code to your path, you have to copy and paste that command into your terminal. Then start Claude Code with this command: 
claude
After setting up your preferences, you’ll be prompted to log in. Select Option 2, which will open a browser tab. Use your Gmail account to log in. When done, you can quit Claude Code. Finally, run this command in your terminal (not inside Claude Code), replacing the <...> placeholders: 
claude mcp add --transport http <MCP_Name> <Host> \\
  --header "X-API-Key: <Key>" \\
  --header "X-API-Secret: <Secret>" \\
  --header "X-API-User: <User_email>"
The convention for naming your MCP server is: upflow-api-<Org_Name> (if you want to setup the MCP for a single org you can simply use upflow-api).

​
Step 3 – Enjoy ✨

To start Claude Code, run this command: 
claude
Then verify your MCP connection with this command (inside Claude Code): 
/mcp
You should see your MCP server with status 🟢 connected. Here are some prompt examples to try:
  • 🔮 Retrieve top debtors 
    Who are the top 5 debtors by outstanding balance?
  • 💸 How much we collected during a period 
    How much have we collected since the beginning of the month?
  • ☎️ Get main contacts 
    Who is the main contact of X?

​
🛠️ Troubleshoot

  • How to verify the MCP entry in Claude configuration Open the file: /Users/<your_username>/.claude.json and check for the mcpServers entry.
    • If the file doesn’t exist, verify your Claude Code installation.
    • If there’s no mcpServer entry, retry Step 3.
  • MCP server already exists If you try to add the MCP server using the command in Step 3 (Claude CLI) and encounter an error, you may have already added it previously. To fix this, remove the existing MCP server and add it again. First, list your MCP servers with this terminal command: 
    claude mcp list
    If you see an MCP server with the same name, delete it with: 
    claude mcp remove <mcp_name>
    Then run the command from Step 2 to add the MCP server.